| Most important economical traits in crop breeding are quantitative traits.Exploring their genetic basis,mapping associated genes and predicting hybrids' performance would help to purposefully reform the genetic architecture of crops and improve crop breeding.Thus,quantitative genetics analysis is essential for modern crop breeding.Quantitative genetics usually aims at simple bi-parental segregation population,such as F2 population,which captures a narrow genetic basis and little favorable alleles.Incorporating multi-parental breeding populations in the study,such as mating design populations is therefore needful.However,this approach has several limitations.1)Relatively little has been known about the application of QTL mapping in North Carolina(NC)design II.Especially,epistatic mapping method in NCII population is still mysterious.2)Previous studies about genomic selection were used to predict the performance of individuals from inbred lines or bi-parental lines,and non-additive effects have always been ignored in the prediction models.Therefore,it is necessary to explore the application of genomic selection method in NCII population.3)Now,there are some arguments towards statistical methods for nested association mapping(NAM).For instance,which is better between single family analysis and joint multiple-family analysis?How to treat the effects of one QTL in different sub-populations?Therefore,we addressed these problems from three aspects.Firstly and foremost,an interacted QTL mapping method was developed for NCII mating design.The mapping results were used to calculate breeding values of all the missing F1 hybrids,and then to predict elite parents and hybrids combinations.Secondly,a genomic selection approach(full model-based GBLUP)was applied to the NCII mating design for hybrid prediction.Lastly,following a joint multiple-family analysis,a multiple QTL mapping method which differentiated the effects of QTL in different families was developed.The main results were as follows:1)An interacted QTL mapping model,including all potential main and interacted effects,in a NCII mating design population was developed.Empirical Bayes method could estimate all effects in the model when the number of effects is not very large.But when the number of effects in one model is huge,combining Bulked Segregant Analysis(SA)and Empirical Bayes could handle a high throughput data and save the computing time.Simulation results are as follow:with the sample size of 725,the power in the detection of QTL mapping was 46.75%,92.88%and 99.25%,respectively,for heritabilities of 0.02,0.05 and 0.08;with the heritability of 0.05,the power was 64.25%,76.88%and 87%,respectively,for sample size of 400,600 and 800;false positive rate stayed around 0.2‰;absolute bias between estimated and true effects of QTL was less than 0.03.New method analyzed the rapeseed oil content data in a partial NCII mating design.8 main QTL and 37 interacted QTL were identified and explained 60.41%of phenotypic variance.Among these QTL,3 main QTL linked with maker BnGMS488A,BRAS078A and O110-C01,as well as one interacted QTL MR119dxNa12-A02B were identified by other four association mapping methods;4 main QTL and 26 interacted QTL were validated with more than one other method.Besides,16 main or interacted QTL were the same with previous studies for mapping oil content QTL,and 35 associated markers were linked with those in previous studies.A high correlation(0.76)between the predicted and observed phenotypes was observed,then 10 elite restorer lines,10 elite sterile lines and 10 elite parental crosses were predicted.2)GBLUP method for genomic selection was applied to predict the elite parental lines and cross combinations in NCII mating design.The genetic effect was partitioned into six components:additive,dominance,additive-by-additive,additive-by-dominance,dominance-by-additive and dominance-by-dominance.Genome-wide makers were used to estimate the genomic relationship among individuals,and then predict the phenotypic values of missing F1.The rapeseed oleic acid content in partial NCII mating design was analyzed.The results showed that the genetic variances of additive,additive-by-additive and additive-by-dominance,were 6.97,15.26 and 12.45,respectively;explained 16.58%,36.31%and 29.62%of phenotypic variance.The estimated variance components were used to predict the phenotypic values,and the goodness of fit was 93.91%.Cross validation experiments indicated a relatively good predictability for phenotypic values(average R2 is 49.27%;average r2 is 0.4949).A series of Monte Carlo simulation based on the real NAM genotypic data validated our new method.Then,all F1 with observed and predicted phenotype values were used to analyze combining ability and predict 10 elite restorer lines,10 elite sterile lines and 10 elite parental crosses.3)At first single family composite interval mapping(CIM)for three flowering time traits in maize was conducted.137,138 and 89 QTL were identified by CIM,and explained 66.4%,74.6%and 94.4%of phenotypic variance,respectively.The CIM results validated our hypothesis that one QTL might have different effects in different sub-populations.Then followed a joint multiple-family analysis method,we developed a multiple QTL model which differentiated the effects of QTL in different family.The new method re-analyzed the flowering time traits in maize.77,19 and 15 QTL were identified for each trait,respectively,and explained 90.11%,89.44%and 82.5%,which cover 93%,96%and 91%of the QTL from CIM,respectively.In addition,it detected 25,29 and 32 new QTL with minor effects,respectively,for three traits.Moreover,a series of Monte Carlo simulation based on the real NAM genotypic data validated our new method.When the sample size retained 964,average QTL power was 59.2%,81%and 91.1%,respectively,for heritability of 0.03,0.05 and 0.08.When the heritability of each QTL retained 0.07,average QTL power was 64.8%,91.7%and 95.1%for sample size of 400,600 and 800,respectively.The bias of true and estimated effects ranged from-0.068 to 0.040.In conclusion,we suggested one quantitative trait analysis method to assist the hybrid breeding.At first,QTL mapping and association analysis was proposed to detect favorable alleles,then genomic prediction method was suggested to predict the potential hybrid performance in hybrid breeding,and further provide evidence for breeding by design. |
PDF Full Text Request